Artificial Intelligence in Cyber Security:Generalized Model Approach to Accurate Fraud Detection in Credit Card

To enhance cyber security in banking systems, we have developed a generalizable machine learning model for accurate fraud detection using the European card benchmark dataset. In this research, we addressed high-class imbalanced data through PCA-based comparative analysis to achieve efficient results. Subsequently, to increase fraud detection performance, a simple Multilayer Perceptron, PCA-based Random Forest, and standard Random Forest classifier models were developed with variations in hidden layers, learning rates, and epochs, as well as fine-tuning of parameters. The cross-validation technique was used to evaluate the research study and mitigate the overfitting problem. Finally, the Random Forest classifier demonstrated the highest performance compared to existing work, achieving 100% accuracy and 99.99% precision without overfitting. The optimized RF model provides a scalable solution for safeguarding sensitive transactions and mitigating fraudulent activities in banking systems.